parthian

The R package parthian for landscape connectivity analysis

https://github.com/emilio-berti/parthian

Repository

The R package parthian for landscape connectivity analysis

README.Rmd

---
author: Emilio Berti
output: 
  github_document:
    pandoc_args: --webtex
---

[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.10890031.svg)](https://doi.org/10.5281/zenodo.10890031)

```{r, include = FALSE}
knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  fig.width = 4,
  fig.height = 4,
  fig.align='center',
  dpi = 300,
  out.width = "50%"
)
```

# Development and dependecies

The R package _parthian_ is developed and maintained by Emilio Berti ().
There are several dependencies for _parthian_:

  - Rcpp
  - igraph
  - terra
  - enerscape
  
They are all stable packages with long history, except for _enerscape_, which I developed in 2021 and maintain since then:  and .

```{r}
library(terra)
library(enerscape)
library(igraph)
library(parthian)
```

# Workflow

## Introduction

The scope of _parthian_ is to quantify the importance of areas in the landscape based on energy cost of movement for animals.
This is achieved by building a weighted graph between adjacent cells using as weights the cost of transport ($E_{COT}$) between them.
This weighted graph is used to obtain least-cost paths and to rank areas based on their importance in promoting movement across such paths.

There are two datasets in _parthian_:

  - dem: a digital elevation model for an area in Sicily, Italy.
  - pa: the protected areas in the same region.

These are matrices, as it is easier to store them in an R package.
The first thing is to transform them into raster.

```{r torasters}
data(dem)
dem <- rast(
  dem,
  crs = "+proj=utm +zone=32 +datum=WGS84 +units=m +no_defs"
)
data(pa)
pa <- rast(
  pa,
  crs = "+proj=utm +zone=32 +datum=WGS84 +units=m +no_defs"
)
plot(dem, col = colorRampPalette(c("darkblue", "seagreen", "white"))(100))
plot(pa, add = TRUE, col = adjustcolor("gold", alpha.f = .5), legend = FALSE)
lines(as.polygons(pa))
```

The resolution and extent of the layers are wrong (I need to fix this in the package data), but it does not matter too much for the examples.

## Energy landscape

The next step is to calcualte the energy landscape for the animal.
Here, I am assuming an animal of 10 kg.

```{r enerscape}
en <- enerscape(dem, 10, "kcal")
plot(en, col = colorRampPalette(c("grey95", "tomato", "darkred"))(100))
plot(pa, add = TRUE, col = adjustcolor("gold", alpha.f = .5), legend = FALSE)
lines(as.polygons(pa))
```


## Weighted graph 

The main task of _parthian_ is to create a graph where vertices (_V_) are the cells of the energy landscapes and weighted edges (_E_) $E_{ij} = E_{COT}$ if two cells are adjacent, and $E_{ij} = 0$ if they are not.

```{r cost-graph}
g <- cost_graph(en)
```

Generally, there are as many vertices as number of cells

```{r vertices}
length(V(g)) == ncell(en)
```

but the number of edges may be lower than $8V$, as some paths may be blocked, in this example by the sea.

```{r edges}
length(E(g)) == ncell(en) * 8
```

## Least cost paths
Least-cost paths can be obtained using the weighted graph created by `cost_graph()` and the _igraph_ `shortest_paths()` function.
First, let's get the centroids of the protected area, after exlcuding very small areas ($\leq 100 m^2$):

```{r centroids}
pas <- disagg(as.polygons(pa))
pas <- pas[expanse(pas, "m") > 100, ]
centrs <- centroids(pas)
plot(pa, col = "gold")
points(centrs, cex = 1, pch = 21)
points(centrs[c(1, 4), ], cex = 1, pch = 20)
```
Let's calcualte the least-cost path between the two areas highlighted by the solid circle.
Because there is a one-to-one correspondence between cell and vertex ID, this can be achieved by:

```{r lcp}
xy <- extract(en, centrs[c(1, 4), ], cells = TRUE)
lcp <- shortest_paths(g, xy$cell[1], xy$cell[2])
path <- lcp$vpath[[1]]
path <- xyFromCell(en, as.numeric(path))
path <- vect(path, crs = crs(dem))
total_costs <- sum(extract(en, path)[["EnergyScape"]])
```

```{r plot-path}
plot(en, col = colorRampPalette(c("grey95", "tomato", "darkred"))(100))
plot(pa, add = TRUE, col = adjustcolor("gold", alpha.f = .5), legend = FALSE)
lines(as.polygons(pa))
lines(as.lines(path), lw = 3, col = "green4")
text(220, 350, paste("Energy costs:", round(total_costs), "kcal"))
```

The function `parthian_path()` wraps the above code and can be called from _parthian_.

```{r path}
lcp <- parthian_path(g, en, centrs[1], centrs[4])
lcp
```

`parthian_path()` returns the least-cost path as a SpatVector and its total travel costs, which are the same as before.

```{r replot-path}
plot(en, col = colorRampPalette(c("grey95", "tomato", "darkred"))(100))
plot(pa, add = TRUE, col = adjustcolor("gold", alpha.f = .5), legend = FALSE)
lines(as.polygons(pa))
lines(lcp$lcp, lw = 3, col = "green4")
text(220, 350, paste("Energy costs:", round(lcp$costs), "kcal"))
```

Instead of calculating least-cost paths manually, _parthian_ uses the function `parthian_paths()` to obtain them iteratively between all cells.

```{r paths}
lcps <- parthian_paths(g, en, centrs)
lcps
```
The output of `parthian_paths()` is a list with:

  1. _lcps_: the lines of the least-cost paths (SpatVect).
  2. _costs_: the matrix with energy costs between cells, symmetric.

```{r paths-plot}
plot(en, col = colorRampPalette(c("grey95", "tomato", "darkred"))(100))
plot(pa, add = TRUE, col = adjustcolor("gold", alpha.f = .5), legend = FALSE)
lines(pas)
lines(lcps$lcps, lw = 3, col = "green4")
```

As a rule of thumb, if you want to call `parthian_path()` several times, the usage of `parthian_paths()` is preferred.

Owner

Citation (CITATION.cff)

cff-version: 1.2.0
message: "If you use this software, please cite it as below."
authors:
  - family-names: Berti
    given-names: Emilio
    orcid: https://orcid.org/0000-0001-9286-011X
title: parthian
version: 2.0.4
identifiers:
  - type: doi
    value: 10.5281/zenodo.10890031

date-released: 2024-03-28

GitHub Events

Issues and Pull Requests

Last synced: 12 months ago